Anthropic’s Claude Haiku 4.5 matches May’s frontier model at fraction of cost

On Wednesday, Anthropic introduced Claude Haiku 4.5, an innovative AI language model that reportedly matches the performance of its earlier frontier model, Claude Sonnet 4, which was released five months ago. Remarkably, Haiku 4.5 achieves this at only a third of the cost and boasts over double the processing speed. This new model is now accessible to all users of the Claude app, web, and API services. If the performance benchmarks provided by Anthropic are validated through independent testing, it showcases the company's ability to deliver a highly efficient coding model that rivals its more advanced counterparts, including GPT-5, while significantly reducing both cost and operational speed. Anthropic offers the Claude family in three distinct sizes: Haiku (small), Sonnet (medium), and Opus (large). Larger models tend to utilize more extensive neural networks, providing deeper contextual understanding but at the expense of higher costs and slower processing times. Thanks to a technique known as distillation, Anthropic has succeeded in developing smaller models that can perform at levels comparable to older, more complex models, especially for specific tasks like coding. However, this often results in a trade-off regarding the breadth of stored knowledge. For users seeking detailed discussions on complex subjects such as foreign policy or historical analysis, larger models like Sonnet or Opus may be preferable, despite their potential inaccuracies. Conversely, for those needing rapid coding support that prioritizes the translation of concepts, Haiku proves to be a more suitable and economical choice.